Related Experiment Video
Updated: Sep 10, 2025
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
681
Enhancing Clinical Decision Support with Adaptive Iterative Self-Query Retrieval for Retrieval-Augmented Large
Srinivasagam Prabha1, Cesar A Gomez-Cabello1, Syed Ali Haider1
1Division of Plastic Surgery, Mayo Clinic, 4500 San Pablo Road, Jacksonville, FL 32224, USA.
Bioengineering (Basel, Switzerland)
|August 28, 2025
Summary
The Self-Query Retrieval (SQR) framework improves clinical decision support by automatically structuring and refining questions for large language models (LLMs). This enhances the accuracy and relevance of AI-generated medical guidance for physicians.
Area of Science:
- Artificial Intelligence in Medicine
- Clinical Decision Support Systems
- Natural Language Processing
Background:
- Retrieval-Augmented Generation (RAG) using large language models (LLMs) shows potential for clinical guidance but is limited by query quality.
- Physician cognitive load can be reduced by AI, but requires accurate and structured information retrieval.
- Current RAG systems struggle with unstructured or ambiguous clinical queries, impacting reliability.
Purpose of the Study:
- To introduce and evaluate the Self-Query Retrieval (SQR) framework for enhancing clinical question clarity and structure.
- To improve the accuracy, relevance, and retrieval quality of LLM-generated clinical guidance.
- To assess the effectiveness of automated query refinement modules (PICOT, SPICE, IQR) within the SQR framework.
Main Methods:
- Developed the adaptive SQR framework integrating PICOT, SPICE, and Iterative Query Refinement (IQR) modules.
- Implemented SQR on the Gemini-1.0 Pro LLM and benchmarked with 30 postoperative rhinoplasty queries.
- Evaluated response accuracy and relevance using a Likert scale and retrieval metrics (precision, recall, F1 score).
Main Results:
- The full SQR pipeline achieved 87% accuracy and 100% relevance, significantly outperforming a non-refined RAG baseline (50% accuracy, 80% relevance).
- SQR improved precision, recall, and F1 scores from 0.17, 0.39, 0.24 to 0.53, 1.00, 0.70, respectively.
- PICOT-only and SPICE-only modules showed intermediate improvements, highlighting the benefit of the full pipeline.
Conclusions:
- Automated query structuring and iterative refinement via SQR substantially enhance LLM-based clinical decision support.
- The SQR framework demonstrates significant improvements in accuracy and relevance for clinical guidance.
- SQR's model-agnostic design allows for broad applicability across medical specialties and data sources.
Keywords:
clinical decision supportdecision support systemslarge language modelsretrieval-augmented generationself-query retrievalMore Related Videos
Related Concept Videos
01:12Retrieval
169
Retrieval is the process of getting information out of memory storage and back into conscious awareness. This ability is essential for daily tasks like brushing hair and teeth, driving to work, and performing job duties. Retrieval occurs in three ways: recall, recognition, and relearning.
Recall involves accessing information without cues, such as during an essay test, where individuals must retrieve facts and concepts from memory unaided. Another example is remembering the name of a colleague...
Recall involves accessing information without cues, such as during an essay test, where individuals must retrieve facts and concepts from memory unaided. Another example is remembering the name of a colleague...
169
01:45ER Retrieval Pathway
3.9K
In the secretory pathway, vesicles transport proteins from one cellular compartment to another in forward transport to deliver the protein to its correct location. Occasionally, misfolded proteins and incorrect proteins escape their original compartments, and a retrieval pathway is used to return the escaped proteins to their original compartment.
The ER uses many checkpoints to prevent the entry of incorrectly folded or a resident protein as cargo onto a transport vesicle. These mechanisms...
The ER uses many checkpoints to prevent the entry of incorrectly folded or a resident protein as cargo onto a transport vesicle. These mechanisms...
3.9K